The Journal of Neuroscience, May 1994, 14(5): 31063121

Neuronal Loss Induced in Limbic Pathways by Kindling: Evidence for Induction of Hippocampal Sclerosis by Repeated Brief Seizures

Jose E. Cavazos,1,2 lndranil Das, and Thomas P. Sutulal ‘The Neuroscience Training Program and the Departments of and Anatomy, University of Wisconsin, Madison, Wisconsin 53792 and *Duke University Medical Center, Durham, North Carolina 27710

Repeated kindled seizures induce long-lasting physiological mechanisms in the neuronal damage associated with status epi- and morphological alterations in the . lepticus, a severe, sometimes life-threatening condition char- In the (DG), the morphological alterations in- acterized by prolonged or repetitive seizures that are not fol- duced by’kindled seizures include loss of polymorphic neu- lowed by return to normal consciousness. Status epilepticus is rons in the hilus, mossy fiber axon sprouting, and synaptic often followed by extensive neuronal damage involving the hip- reorganization of the mossy fiber pathway. In this study, pocampus, other limbic structures, neocortex, and subcortical quantitative stereological methods were used to determine areas, and by cognitive deficits (Hauser, 1983). the distribution and time course of neuronal loss induced by Excitotoxic damage has been demonstrated in several exper- 3, 30, or 150 kindled generalized tonic-clonic seizures in imental models of status epilepticus. In primates, prolonged hippocampal, limbic, and neocottical pathways. Neuronal loss seizures induce neural damage despite maintenance of systemic was observed in the hilus of the DG and CA1 after three metabolic parameters in the normal range (Meldrum et al., 1973). generalized tonic-clonic seizures, and progressed in these In rodents, intense electrographic seizure activity evoked by sites to 49% and 44% of controls after 150 seizures. Neu- repeated frequent stimulation of the from the ronal loss was also observed in CA3, entorhinal cortex, and entorhinal cortex (EC) to dentate gyrus (DG) (10 set of 20 Hz the rostra1 endopyriform nucleus after 30 seizures, and was stimulation per minute for 24 hr) also induces damage to neu- detected in the layer and CA2 after 150 seizures. rons in the hilus of the DG (Sloviter, 1983, 1987). In another There was no evidence of neuronal loss in the somatosen- rodent model, administration of kainic acid, an analog of the sory cortex after 150 seizures. The time course of the neu- excitatory amino acid L-glutamate, induces intense acute seizure ronal loss demonstrated selective vulnerability of hippocam- activity, which is accompanied by neuronal damage in CAl, pal neuronal populations to seizure-induced injury, and CA3, the hilus of the DG, and other limbic structures (Nadler suggests that even brief seizures may induce excitotoxic and Cuthbertson, 1980; Ben-Ari, 1985). Brief treatment with injury in vulnerable neuronal populations. Repeated brief sei- phenobarbital, which suppresses the acute seizure activity in zures induced neuronal loss in a distribution that resembled this model, prevents excitotoxic neuronal damage in the hilus hippocampal sclerosis, the most common lesion observed ofthe DG, reduces associated mossy fiber sprouting, and reduces in human . The results demonstrated that kindling enhanced susceptibility to evoked seizures (Sutula et al., 1992). induces alterations in neural circuitry in a variety of locations Although it is generally believed that brief sporadic seizures in the limbic system, and suggest that hippocampal sclerosis do not induce excitotoxic damage, human epilepsy is frequently may be acquired in human epilepsy as a consequence of associated with neuronal loss and gliosis in CA 1, CA3, and the repeated seizures. hilus of the DG. This pattern of neuronal loss, which may vari- [Key words: , dentate gyrus, entorhinal cor- ably involve other hippocampal areas, is referred to as hippo- tex, endopyriform nucleus, epilepsy, seizures, kindling, ex- campal sclerosis, Ammon’s horn sclerosis, or mesial temporal cifotoxicity, cell death, ] sclerosis, and is the most common neuropathological finding in human (Babb and Brown, 1987; Gloor, Excessive neuronal activation by the excitatory amino acid 199 1). For over a century, there has been controversy about transmitter glutamate induces neuronal injury and death that hippocampal sclerosis as a cause or effect of the repeated, brief has been referred to as excitotoxicity (Olney et al., 1974; Nadler seizures that are a feature of temporal lobe epilepsy (Babb and and Cuthbertson, 1980). There has been particular interest in Brown, 1987; Sutula, 1990; Gloor, 199 1). the molecular mechanisms of excitotoxicity, as excitotoxic dam- The distribution of neuronal loss in human hippocampal scle- age may play a role in human disorders such as , head rosis resembles the pattern ofexcitotoxic neuronal loss observed trauma, and epilepsy (Choi and Rothman, 1990). in models of status epilepticus (Sloviter and Damiano, 1981; Clinical and experimental studies have implicated excitotoxic Ben-Ari, 1985). It was of interest to determine whether repeated brief seizures, which are not accompanied by the systemic phys- iological and metabolic disturbances encountered in status epi- Received June 24, 1993; revised Sept. 15, 1993; accepted Nov. 1, 1993. lepticus, might induce neuronal loss. In a previous study, we This research was supported by the Klingenstein Fund and the NINDS. We observed that brief seizures evoked by kindling, a model for the thank L. Haberly and R. Kalil for their helpful comments. Correspondence should be addressed to Thomas P. Sutula, M.D., Ph.D., De- sporadic seizures that occur in human temporal lobe epilepsy partment of Neurology H4/6 14, University of Wisconsin, Madison, WI 53792. (Goddard et al., 1969; McNamara, 1986), induced progressive Copyright 0 1994 Society for Neuroscience 0270-6474/94/143106-16$05.00/O neuronal loss in the hilus of the DG (Cavazos and Sutula, 1990). The Journal of Neuroscience, May 1994, 14(5) 3107

In the present study, we report that repeated kindled seizures (2) In a horizontal section about 1000 wrn ventral from the most induced progressive neuronal loss in the hippocampal formation dorsal hippocampus (Fig. 2), neuronal density was measured in the stratum pyramidale of the (CA3c, CA3a, CA2, in a pattern that closely resembles hippocampal sclerosis. Pro- CAlc. and CAla). lavers 2 and 6 of the medial EC (MEC2. MEC6). gressive neuronal loss was also observed in other limbic areas and layers 2 and 6’ofthe lateral EC (LECZ, LEC6). At th’is septotemporal implicated in epilepsy, including the EC and endopyriform nu- level, pyramidal in CA3c are partially enclosed by granule cells cleus (EN), but was not detected in the somatosensory area of in the stratum granulosum. (3) In a horizontal section about 2800 pm ventral from the most the neocortex. dorsal hippocampus that included the motor nuclei of cranial nerves Preliminary observations have appeared in abstract form III and IV (Fig. 3) neuronal density was measured in the hilus of the (Cavazos and Sutula, 199 1). DG, the stratum pyramidale of the hippocampus proper (CA3c, CA3a, CA2, CAlc, and CAla), layers 2 and 6 ofthe medial EC(MEC2, MEC6), layers 2 and 6 of the lateral EC (LEC2, LEC6), the stratum granulosum Materials and Methods of the DG, and the rostra1 EN. In a few brains that were sectioned at a slight angle to the true horizontal plane, the EN was not present in this Kindling and surgical procedures standardized section, as the EN is quite anterior to the motor nuclei of Sprague-Dawley male rats (250-350 gm) were anesthetized with pen- cranial nerves III, IV, and ventral hippocampus. In these cases, it was tobarbital (60 mg/kg, i.p.) and stereotaxically implanted with an insu- necessary to select a more ventral section to measure appropriately the lated stainless steel bipolar electrode for stimulation and recording. The neuronal density of the EN. electrode was implanted in the perforant path near the region of the Quantitative counting methods. Accurate neuronal counting may be angular bundle (8.1 mm posterior, 4.4 mm lateral, 3.5 mm ventral from difficult due to errors introduced by double counting of neurons cut at bregma), in the (1.5 mm posterior, 4.2 mm lateral, 8.8 mm the edges of sections, or systematic changes in tissue volume. Although ventral from bregma), or in the (9 mm anterior, 1.2 mm stereological theory suggests that the most accurate neuronal densities lateral, 1.8 mm ventral from bregma). After a recovery period of 2 can be obtained by counting nucleoli (Konigsmark, 1969) previous weeks, the unrestrained awake animals received twice daily kindling studies have revealed that approximately 40% of polymorphic neurons stimulation (5 d per week) with a 1 set train of 62 Hz biphasic constant- in the hilus of the DG in the rat contain multiple nucleoli (Cavazos and current 1 .O msec square wave pulses. The stimulation was delivered at Sutula, 1990). As counting nucleoli would thus result in falsely high the lowest intensity that evoked an afterdischarge according to standard neuronal densities, neurons were counted only if the nucleus could be procedures (Cavazos et al., 199 1). The electroencephalogram and af- identified in the section. Nuclear counts were obtained ipsilateral and terdischarges were recorded from the bipolar electrode, which could be contralateral to the stimulating electrode at the standardized locations switched to the stimulator by a digital circuit for the delivery of the identified by the criteria described in the previous paragraph, by manual kindling stimulation. The evoked behavioral seizures were classified counting with an eyepiece reticule (25 mmz). The investigator perform- according to standard criteria (Racine, 1972; Sutula and Steward, 1986). ing the counts was unaware of the identity of the sections, and the order Rats received stimulation until three class V seizures were evoked. of examination of the sections was randomized. In each reticule field, Another group received kindling stimulation until 30 class V seizures nuclei that overlapped the lower and left edges of the reticule were were evoked. A third group received kindling stimulation until 150 class counted, but nuclei that overlapped the upper and right edges were not V seizures were evoked. Control groups were normal age-matched rats counted. In each location, the number of neurons and their relative and unstimulated rats implanted with electrodes for 3 months. position were recorded using a camera lucida, and were summed to obtain the nuclear count per reticule field. The magnification selected Histological procedures to obtain the nuclear count per reticule field at each location was selected to include lo-40 neurons as listed in Table 1. In the hilus and stratum At least 18 hr after the last seizure or after the period of control electrode granulosum of the DG, and the rostra1 EN, two reticule fields were implantation, the rats were deeply anesthetized and perfused transcar- selected for each ipsilateral and contralateral location as indicated in dially with an aqueous solution of 10% (v/v) formalin in 0.9% (w/v) Figures 1 and 3. In pilot experiments, the nuclear counting procedure NaCl for lo-20 min. The brains were removed, postfixed for at least in this study had an interobserver variability of less than 10%; the 10 d in same solution, dehydrated for 35 d in graded concentrations of intraobserver variability was 3% (Cavazos and Sutula, 1990). Previous alcohols, and then embedded in graded concentrations of low-viscosity studies have revealed that these methods are quite sensitive for assess- nitrocellulose (celloidin) for 50 d. Embedding in low-viscosity nitro- ment of neuronal loss, which typically exceeds 15-20% before a lesion cellulose provides superior preservation of the neuronal architecture, is reliably detected by visual inspection (Konigsmark, 1969; Cavazos and is particularly suitable for quantitative studies (Konigsmark, 1969). and Sutula, 1990). Horizontal 20 pm sections were cut using a sliding microtome from the Mean neuronal density (N,) for each counted region was calculated surface of neocortex throughout the most ventral hippocampus. All from the average nuclear count per reticule field (n,) obtained from the sections were retained and stored in 70% ethanol. Every fifth section septal and temporal sections as defined in Figures l-3, according to the was stained with cresyl violet, but additional sections were also stained Floderus formula: N, = n, * t/(t + d - 26), where N, is the corrected to ensure that equivalent areas were available for quantitative assess- neuronal density, n, is the average nuclear count per reticule field, ment in each specimen. In a pilot study, this histological method resulted t is the section thickness, d is the average nuclear diameter at each of the in about 40% brain shrinkage compared to the volume calculated im- mediately after perfusion; the amount oftissue shrinkage with this meth- counted regions, and b is a constant that represents the limit of optical resolution for each objective (Konigsmark, 1969; Cavazos and Sutula, od is comparable to previous studies (Amaral, 1978). 1990). For this study, the constant b for each objective was as follows: 20x (NA. 0.75) b = 0.36 wm; 40x (NA 0.95), b = 0.28 pm; 100x oil counting procedures immersion (NA 0.19) b = 0.19 pm (Konigsmark, 1969). For each Locations for neuronal counting. Neuronal counts were obtained in the section, the thickness (t) of each section was also determined by mea- dentate gyrus (DG), hippocampus, entorhinal cortex (EC), the rostra1 suring the distance between upper and lower focal planes of the section endopyriform nucleus (EN), and somatosensory cortex (SSC) in hori- with the stage micrometer using the 100 x oil-immersion objective. The zontal sections at three standardized locations along the septotemporal average nuclear diameter (d) for each cell population is provided in (dorsoventral) axis of the hippocampal formation. The location of the Table 1. As the ratio of section thickness (t) to nuclear diameter (d) was regions assessed by stereological counting methods was identified by the always above 1.5 (range, 1.82-3. lo), the stereological assumptions of following criteria. this method are valid (Clarke, 1992). (1) In a horizontal section about 800 pm ventral from the most dorsal Volume measurements. As systematic changes in tissue volume could hippocampus (Fig. l), neuronal density was measured in the hilus of alter the neuronal density, estimates of the volume of the hippocampal the DG, the stratum granulosum of the DG, and the SSC. At this region (hippocampus proper, , and DG) were obtained ipsi- septotemporal level, the polymorphic neurons in the hilus of the DG lateral and contralateral to the stimulating electrode using a stereological are enclosed in an oval ring formed by granule cells in the stratum approximation method (West et al., 1978; Cavazos and Sutula, 1990). granulosum. The neuronal density in the hilus of the DG was determined The volume was estimated by measuring the area of the hippocampal at this location by quantitative counting methods in our previous report region of four horizontal sections (1000, 2000, 3000, and 4000 pm (Cavazos and Sutula, 1990). ventral to the most dorsal hippocampal section) using an image analysis 3108 Cavazos et al. * Neuronal Loss induced by Kindling

Figure 1. Representative horizontal cresyl violet-stained section of the dorsal hippocampal formation and adjacent cortex of a normal rat about 800 pm ventral from the most dorsal hippocampus. At this septotemporal level, the polymorphic neurons in the hilus of the DG are enclosed in an oval ring formed by granule cells in the stratum granulosum. Neuronal density was measured at this location in the hilus of the DG, the stratum granulosum of the DG, and in layers 2 and 6 of the SK, at sites indicated by the asterisks. H, hilus of the DG, DC, stratum granulosum of the DG, SSC 2, layer 2 of the somatosensory cortex; SSC 6, layer 6 of the somatosensory cortex. Scale bar, 1000 pm. The Journal of Neuroscience, May 1994, 14(5) 3109

Figure 2. Rcprescntative horizontal cresyl violet-stained section of the dorsal hippocampal formation and adjacent cortex of a normal rat about 1000 pm ventral from the most dorsal hippocampus. At this septotemporal level, pyramidal neurons in CA3c are partially enclosed by granule cells in the stratum granulosum. Neuronal density was measured at this location in CA3c, CA3a, CA2, CAlc, CAla, and layers 2 and 6 of both the medial and lateral EC, at sites indicated by the aslerisks. MECZ, layer 2 of the medial entorhinal cortex; MEC6, layer 6 of the medial entorhinal cortex; LECZ, layer 2 of the lateral entorhinal cortex; LEC6, layer 6 of the lateral entorhinal cortex. Scale bar, 1000 brn. 3110 Cavazos et al. * Neuronal Loss Induced by Kindling

Figure 3. Representative horizontal cresyl violet-stained sections of the ventral hippocampal formation, adjacent cortex, and the region of the rostra1 EN of a normal rat about 3600 pm ventral from the most dorsal hippocampus. The orientation of the two photomicrographs is indicated in the schematic diagram at the upper left. At this level the motor nuclei of cranial nerves III and IV are present. Neuronal density was measured in the hilus of the DG, the stratum pyramidale of CA3c, CA3a, CA2, CAlc, and CAla, layers 2 and 6 of both the media1 and lateral EC, the stratum granulosum of the DG, and the region of the rostra1 EN, at sites indicated by the asterisks. EN, endopyriform nucleus; lot, lateral ; pc, pyriform cortex; III, nucleus of crania1 n. III; IV, nucleus of crania1 n. IV, other abbreviations are as in Figures 1 and 2. Scale bar, 1000 pm. The Journal of Neuroscience, May 1994, 74(5) 3111

Table 1. Optical and measurement parameters

Objec- tive Reti- (plan cule Total apo- Nuclear field Magnifi- chro- diameter (mm? cation matic) (pm) Hilus of DG 0.0256 312.5~ 20x 9.77 + 0.49 Granule layer of DG 0.0006 2000 x 100x 6.46 k 0.24 CA3c 0.0016 1250x 100x 9.69 k 0.27 CA3a 0.0016 1250x 100x 10.99 + 0.25 CA2 0.0016 1250x 100x 10.46 + 0.61 CAlc 0.0016 1250x 100x 7.61 + 0.16 CAla 0.0016 1250x 100x 8.71 k 0.10 MEC2 0.0025 1000x 40x 9.51 * 0.04 MEC6 0.0025 1000x 40x 6.84 k 0.39 LEC2 0.0025 1000x 40x 9.29 k 0.11 LEC6 0.0025 1000x 40x 1.42 k 0.01 ssc2 0.0025 1000x 40x 8.07 -t 0.21 SSC6 0.0025 1000x 40x 7.11 k 0.21 EN 0.0025 1000x 40x 8.73 + 0.12

system. The formula V= (A, + A,, ,)t/2 was used to estimate the volume between the sections, where V is the volume, A, is the area of the hippocampal region at a given level, A,, , is the area at the next level, and t is the distance between the two levels (West et al., 1978). For each rat, the mean neuronal density (N,) at each location in the hippocampal formation was calculated according to the formula N, = N,* VJ V,, where V, is the estimated hippocampal volume of the given rat, and V, is the estimated mean hippocampal volume of control rats (Cavazos and Su- tula, 1990). Statistical methods. The differences in nuclear counts per reticule field (n,), mean neuronal densities (N,), and volume-corrected neuronal den- sities (NJ for each region were statistically analyzed using a general linear model for multiway analysis of variance (GLM-ANOVA, split plot). Confidence intervals were calculated for each of the significant variables (i.e., number of class V seizures) using a one-way ANOVA (Tukey’s multiple comparison method), and were considered significant when p < 0.05. All measures reported in this study are expressed as mean f SEM.

Results Neuronal densities (NJ were calculated from the average nuclear counts (n,) in 12 control rats and a total of 20 rats that expe- rienced 3, 30, or 150 class V seizures evoked by kindling. As the differences in neuronal density as a function of the site of stimulation were not significant (p = 0.42, one-way ANOVA), rats stimulated in the perforant path, amygdala, or olfactory bulb were combined for further analysis. The group of rats with three class V seizures (N = 5) included four rats that received stimulation in the perforant path and one rat stimulated in the amygdala. The group with 30 class V seizures (N = 6) included five rats that received stimulation in the perforant path and one rat stimulated in the amygdala. The group of rats with 150 class V seizures (N = 9) included three rats that received stimulation in the perforant path, three rats stimulated in the amygdala, and three rats stimulated in the olfactory bulb. The overall results were not changed if statistical analysis was restricted to rats that received only perforant path stimulation. Calculated hippocampal volume The average hippocampal volume of control rats was 20.9 + 0.6 mmj. For the group of rats that experienced three class V Table 3. Morphological parameters

Pyramidal layer of CA2 Pyramidal layer of CAlc Pyramidal layer of CA la Septal Temporal Septal Temporal Septal Temporal n; (neuronsketicule field) Controls ipsi 11.17 f 0.74 12.92 f 0.87 29.75 -t 2.42 37.83 zk 2.82 33.83 k 2.28 30.17 + 2.03 contra 11.50 + 0.67 12.75 f 2.50 30.42 f 2.02 40.25 f 2.50 33.08 -t 1.69 30.58 + 1.91 3 Cl v ipsi 11.80 IL 0.33 14.20 f 0.52 29.80 f 0.95 37.20 -c 3.50 33.80 + 3.04 25.40 k 1.51 contra 11.80 + 1.04 13.00 rf: 1.13 33.40 + 1.43 35.60 + 4.40 31.20 + 1.37 28.80 -t 2.34 30 Cl v ipsi 10.33 + 0.30 12.67 + 0.90 26.67 f 1.15 28.50 k 2.37 26.83 f 1.32 21.67 + 1.52 contra 11.33 f 0.99 13.00 + 1.15 25.67 + 1.79 28.83 -t 3.24 27.50 + 1.07 20.50 f 2.55 15OClV ipsi 11.33 k 1.70 10.67 + 0.74 29.11 + 5.22 23.22 t 1.61 28.44 k 4.39 15.33 + 1.28 contra 11.22 k 1.11 10.11 + 1.32 26.33 + 1.69 23.00 + 1.26 27.67 31 2.61 17.56 + 1.31 N; (neurons/mm3) Controls ipsi 284,414 k 22,206 304,047 f 19,160 844,511 + 62,312 993,272 + 60,624 914,226 k 56,956 764,781 + 52,650 contra 290,041 k 18,695 307,586 f 17,382 864,384 + 59,801 1,083,218 + 54,120 889,039 + 40,809 790,008 + 45,946 3 Cl v ipsi 278,909 k 12,816 312,346 k 17,542 786,964 k 23,804 895,402 + 57,139 845,519 -t 55,238 591,747 + 33,778 contra 279,784 k 21,716 284,311 AZ 17,569 892,693 k 36,456 864,647 t 85,125 795,332 k 29,513 679,054 + 57,415 30 Cl v ipsi 247,030 -t 8,796 293,383 + 17,710 715,960 + 36,212 734,204 + 41,386 684,578 iz 25,864 535,383 f 26,085 contra 252,427 k 19,800 309,564 + 23,064 638,774 + 40,329 765,940 t 65,194 663,880 k 45,073 519,327 + 49,421 150 Cl v ipsi 246,895 k 20,735 263,853 + 15,088 705,769 + 74,222 644,927 + 32,131 668,414 k 63,998 407,601 f 29,086 contra 259,158 k 23,187 244,190 f 24,246 685,366 + 39,545 638,795 f 35,295 685,040 + 56,289 459,340 f 23,938

Table 4. Morphological parameters

Medial entorhinal cortex (layer 2) Medial entorhinal cortex (layer 6) Lateral entorhinal cortex (layer 2) Lateral entorhinal cortex (layer 6) Septal Temporal Septal Temporal Septal Temporal Septal Temporal n; (neurons/reticule field) Controls ipsi 17.67 + 1.63 14.50 I!z 1.31 20.67 + 1.31 25.83 f 3.08 18.58 + 1.33 13.33 k 1.54 18.00 + 1.03 18.42 k 1.94 contra 15.83 f 1.32 14.00 k 0.53 20.25 + 1.07 27.83 f 1.94 20.67 + 1.22 14.92 I~I 1.06 19.58 z!z 1.22 24.17 k 1.61 3 Cl v ipsi 16.60 f 2.38 12.40 k 0.83 21.00 + 2.62 27.40 rk 1.87 20.20 + 2.16 14.00 f 1.36 19.20 k 1.97 20.40 + 1.49 contra 18.40 f 2.36 14.80 k 1.15 22.80 + 0.72 28.00 f 3.36 19.80 + 2.94 15.20 + 2.03 20.00 k 1.62 19.00 + 1.55 30 Cl V ipsi 13.33 f 0.81 12.67 k 1.24 18.17 +- 1.14 28.17 f 2.60 15.67 + 1.07 10.67 f 0.81 16.33 k 0.38 19.17 + 1.96 contra 14.67 k 1.10 12.33 -t 1.07 20.17 + 2.30 22.33 zk 2.00 16.83 + 1.07 10.00 -t 1.00 17.67 k 1.97 15.00 f 1.56 150 Cl V ipsi 13.33 f 1.40 11.00 -t 1.68 16.56 + 2.29 23.33 f 2.21 15.33 + 0.87 10.89 f 2.05 16.89 k 2.50 18.22 + 3.25 contra 17.33 -+ 2.55 11.11 k 1.55 20.33 + 3.43 23.44 f 2.62 17.56 f 1.22 10.00 * 1.13 17.33 k 2.00 16.22 + 1.47 N; (neurons/mm”) Controls ipsi 296,692 k 24,167 230,372 + 22,764 391,874 + 20,098 453,406 k 53,732 315,475 + 19,436 208,690 f 21,757 332,752 + 17,212 315,496 + 32,837 contra 264,962 + 21,396 225,313 f 6,997 387,835 f 26,268 500,127 k 30,024 349,702 k 19,321 239,867 + 13,570 362,462 + 24,659 426,480 t 29,355 3 Cl v ipsi 258,235 f 30,834 180,127 f 9,338 364,671 k 36,294 439,482 k 16,855 314,966 k 23,333 204,333 f 16,566 327,187 + 26,362 321,184 + 18,064 contra 286,642 + 29,257 216,383 + 11,291 406,219 k 9,563 449,339 k 37,402 311,278 k 42,289 223,847 + 24,428 344,429 + 20,803 300,717 t 15,690 30 Cl V ipsi 212,568 + 11,987 196,253 + 18,402 323,979 + 18,050 484,798 + 38,587 252,387 + 17,190 166,237 + 9,887 284,665 f 7,160 321,283 k 27,024 contra 218,649 f 14,339 196,245 k 14,631 327,449 + 23,108 398,151 + 26,226 253,069 + 15,025 160,509 f 12,888 280,825 + 19,355 259,545 + 21,574 150 Cl V ipsi 199,316 t 15,047 178,939 f 21,400 270,583 + 21,063 429,775 f 24,849 236,295 f 15,947 177,102 + 25,777 270,072 + 25,045 323,743 k 45,804 contra 267,182 + 35,595 178,168 k 18,128 346,551 f 52,161 425,596 + 31,439 273,827 + 15,864 163,444 + 13,739 290,035 f 28,052 288,678 -e 16,742 The Journal of Neuroscience, May 1994, 14(5) 3113

Table 5. Morphological parameters

Endopyriform Somatosensory cortex nucleus Layer 2 Layer 6 n; (neurons/reticule field) Controls ipsi 11.96 + 0.55 13.58 k 0.69 16.67 -t 1.83 contra 12.00 * 0.55 14.42 f 0.49 16.42 k 0.59 3 Cl v ipsi 11.80 ? 0.36 14.60 & 1.31 17.40 k 1.22 contra 11.70 & 0.64 15.80 + 1.25 16.20 k 0.91 30 Cl v ipsi 9.83 f 0.63 13.67 + 0.90 14.67 k 1.22 contra 9.25 + 0.56 14.17 + 0.64 14.67 -t 0.45 15OClV ipsi 9.22 t 1.50 15.11 + 1.70 17.22 rt_ 1.73 contra 9.50 k 1.24 14.33 * 1.89 15.33 + 1.54 N; (neurons/mm3) Controls ipsi 196,442 + 10,128 219,265 * 7,738 278,065 + 24,701 contra 198,820 + 8,608 246,503 + 8,170 292,723 + 10,361 3 Cl v ipsi 180,157 I? 5,951 228,124 + 10,865 284,549 + 13,160 contra 178,621 k 8,684 274,477 f 19,528 294,098 k 13,281 30 Cl v ipsi 157,601 -t 7,809 246,578 + 24,276 279,686 + 37,125 contra 154,106 ? 8,572 246,843 + 12,121 266,129 k 6,033 15OClV ipsi 151,986 -I 19,877 256,898 + 22,053 305,784 + 23,463 contra 156,409 + 16,952 251,738 + 26,173 286,033 k 26,901

seizures, the average hippocampal volume was 2 1.3 + 1.O mm3. Significant differences in neuronal density along the septotem- For the group that experienced 30 class V seizures, the average poral axis were not noted in CA3c, CA 1c, CA2, the granule cell hippocampal volume was 22.1 ? 0.8 mm3. After 150 class V layer of the DG, or LEC6. seizures, the average hippocampal volume was 22.6 + 1.5 mm3. Although there was a small increase in the calculated hippo- campal volumes of these kindled rats perfused at 18 hr after the Decreases in neuronal density (NJ as a function of evoked last seizure, this trend did not achieve statistical significance (p seizures = 0.58, one-way ANOVA). As the results using volume-cor- There were progressive decreases in neuronal density in a variety rected neuronal densities (NJ did not differ significantly from of limbic structures as a function of increasing numbers of evoked the results using N, , the overall analysis was performed using seizures. With increasing numbers of evoked seizures, the neu- n, and N,. The morphological parameters n, and N, are provided ronal densities (NJ progressively decreased in the polymorphic in Tables 2-5. region of the hilus of the DG, CAlc, CAla, CA3c, CA3a, the stratum granulosum of the DG, CA2, MEC2, MEC6, LEC2, Septotemporal variation of neuronal densities (NJ in normal LEC6, and the rostra1 EN. Alterations in n, and N, were not rats observed in layer 2 or layer 6 of the SSC (for a summary, see In agreement with previous studies (Amaral et al., 1990), there Table 7, Fig. 4). were differences in the neuronal densities of limbic structures Neuronal loss after three class V seizures. The reduction in nj along the septotemporal axis of the hippocampal formation in and N, was not equivalent in all locations (p < 0.001, GLM- normal rats (for a summary, see Table 6). In normal rats, the ANOVA; see Table 7, Fig. 4). After three class V seizures, there neuronal densities in the hilus of the DG, CA3a, and MEC6 were significant reductions in N, , expressed as percentage of the were significantly higher in the temporal pole than in the septal control value, in the septal (dorsal) pole of the hilus of the DG pole. In contrast, the neuronal densities of CAla, MEC2, and (78%) the temporal hilus of the DG (76%) temporal CAla LEC2 were higher in the septal pole than in the temporal pole. (82%) and temporal CA3a (83%).

Table 6. Septotemporal variation of neuronal densities in normal rats

Neurons/mm’ (k SEMl Sental Temnoral Hilus DG 58,956 + 2,941 72,098 k 4,855 F = 28.59, p < 0.0001 CA3a 260,717 + 17,559 308,624 k 19,394 F = 10.84, p < 0.001 MEC6 391,874 + 20,098 453,406 2 53,732 F = 30.67, p < 0.0001 CAla 914,226 f 56,956 764,781 k 52,650 F = 38.55, p i 0.0001 MEC2 296,692 f 24,167 230,372 -t 22,764 F = 17.28, p < 0.0005 LEC2 315,475 + 19,436 208,690 k 21,757 F = 74.03, p < 0.0001 DG, dentate gyros; MECZ, MEC6, layers 2, 6 of medial entorhinal cortex; LECZ, layers 2,6 of lateral entorhinal cortex. 3114 Cavazos et al. - Neuronal Loss Induced by Kindling

Table 7. Neuronal density as a function of 3,30, or 150 evoked There were no significant differences between the neuronal seizures (expressed as a percentage of normal control) densities obtained ipsilateral and contralateral to the stimulating electrode (Tables 2-5; p = 0.46, GLM-ANOVA). The general Number of results are summarized in Figure 4. class V seizures 3 30 1.50 Septotemporal variations in seizure-induced neuronal loss DG hilus (s) 78% 63% 51% F = 60.75, p < 0.0001 The decreases in neuronal densities tended to be greater in the (t) 76% 61% 54% temporal pole than in the septal pole, as indicated by the ratio CAla 6) - 75% 75% F = 29.39, p < 0.0001 of septal to temporal neuronal density in DG, CAlc, and CAla. (9 82% 68% 56% The ratio of septal:temporal neuronal densities in CAlc signif- CA3a (4 - 83% 84% F = 9.46, p < 0.0005 icantly increased as a function of the number of seizures (p < (t) 83% 82% 78% 0.028, F = 3.42, GLM-ANOVA; Table 3, Fig. 6). This finding CAlc 64 - 79% 81% F = 20.99, p < 0.0001 suggested that the temporal pole of CAlc was more vulnerable 0) - 72% 62% to seizure-induced neuronal loss than the septal pole. CA3c 6) - 73% 77% F = 18.82, p < 0.0001 74% 71% 0) - Discussion CA2 (s) - - - F = 3.38, p < 0.03 (9 - - 83% In this experiment, progressive decreases in neuronal density DG sg (s) - - - F = 5.16, p < 0.006 were observed in the DG, hippocampus, EC, and rostra1 EN @I - 83% 80% after repeated brief kindled seizures. As there were no significant MEC2 6) - 77% 83% F = 4.35, p < 0.011 changes in tissue volume or differential shrinkage in control and 0) - 86% 78% kindled rats, the decreases in neuronal density most likely reflect LEC2 (s) - 76% 77% F = 14.62, p i 0.0005 neuronal loss induced by the repeated brief kindled seizures. @) - 73% 76% Many previous studies have not detected neuronal loss or MEC6 (9 - - 79% F = 4.73, p < 0.008 structural lesions as a consequence of repeated brief kindled (0 - - 90% seizures (Goddard et al., 1969; Crandall et al., 1979; Sutula et LEC6 6) - - 81% F = 8.03, p < 0.001 al., 1988; Represa et al., 1989; Bet-tram and Lothman, 1993). 0) - 78% 83% However, many of these studies did not employ histological and EN - 79% 78% F= 5.21, p < 0.005 quantitative stereological techniques that enabled reliable de- SSC2, SSC6 - - - tection of neuronal loss, or evaluated neuronal populations after only three class V seizures, when significant seizure-induced s, septal; t, temporal; DG, dentate gyrus; DG sg, granule cell layer of the dentate gyrus; MECZ, MEC6, layers 2, 6 of medial entorhinal cortex; LEC2, LEC6, layers neuronal loss first becomes detectable by the methods employed 2, 6 of lateral entorhinal cortex; EN, endopyriform cortex; SSCZ, SSC6, layers 2, in this study. The results of this study do not exclude the pos- 6 of somatosensory cortex. sibility that more subtle neuronal loss might be detected by more sensitive methods after only a few evoked seizures, or a single seizure. Neuronal loss after 30 class V seizures. After 30 class V sei- These experiments demonstrated that repeated electrical zures, there were significant reductions in N, in the septal hilus stimulation that evoked kindled seizures in limbic pathways of the DG (63% ofthe control value), CA3c (73%) CA3a (83%) was sufficient to induce neuronal loss in a variety of limbic areas. CAlc (79%), CAla (75%), MEC2 (77%) and LEC2 (76%). In It seems likely that the neuronal loss was the result of seizures the temporal location, there were significant reductions in N, in rather than damage caused by direct effects of stimulation, as the hilus of the DG (61% of the control value), the stratum there was approximately equal neuronal loss ipsilateral and con- granulosum of the DG (83%), CA3c (740/o), CA3a (82%), CAlc tralateral to the stimulating electrode in all regions selected for (72%) CAla (68%) MEC2 (86%) LEC2 (73%), LEC6 (780/o), study. In support of this conclusion, previous studies have also and the rostra1 EN (79%). After 30 class V seizures, the neuronal demonstrated that kindled seizures rapidly activate hippocam- densities in the hilus of the DG and the pyramidal layer of CA3c pal pathways bilaterally (Stringer and Lothman, 1992), perhaps and CAlc were significantly different from the neuronal den- by the prominent commissural pathways which are a feature of sities in these locations after three class V seizures (one-way the Sprague-Dawley rat. ANOVA, p < 0.05). Neuronal loss after 150 class V seizures. After 150 class V Selective vulnerability of limbic structures to seizure-induced seizures, there were significant reductions in N, in the septal neuronal loss location of the hilus of the DG (5 1% of the control value; Fig. There were progressive seizure-induced decreases in the neu- 5A.B) MEC6 (79%), and LEC6 (8 1%). In the temporal location, ronal density of a variety of limbic structures, but the neuronal there were significant reductions in N, in the hilus of the DG density of the SSC was not altered by as many as 150 seizures. (54% of control value; Fig. 5C,D), the stratum granulosum of These results demonstrated that the quantitative stereological the DG (80%) CA3c (71%), CA3a (78%) CA2 (83%), CAlc methods of this study were not merely detecting a diffuse or (62%; Fig. 6) CAla (56%; Fig. 7) MEC2 (78%) MEC6 (90%) global effect of repeated seizures on central neurons, but rather and the rostra1 EN (78%). After 150 class V seizures, the neu- a specific vulnerability of certain neuronal populations to sei- ronal density in the hilus of the DG, CA3c, CAlc, and CAla zure-induced damage. In particular, there appears to be signif- was significantly different from the neuronal density in these icant vulnerability of the neuronal populations in limbic struc- locations after three class V seizures (one-way ANOVA, p < tures such as the DG, hippocampus, EC, and rostra1 EN. 0.05). These results indicated that neuronal loss in these loca- The present study extended our previous observations about tions was progressive as a function of the number of seizures. the progressive loss of polymorphic neurons in the hilus of the The Journal of Neuroscience, May 1994, 14(5) 3115

HILUS OF THE DENTATE GYRUS DENTATE GRANULE LAYER PYRAMIDAL LAYER OF CA3

PYRAMIDAL LAYER OF CA2 PYRAMIDAL LAYER OF CA1 ENDOPYRIFORM NUCLEUS 120, 120, , 120, /

MEDIAL ENTORHINAL CORTEX LATERAL ENTORHINAL CORTEX SOMATOSENSORY CORTEX 120, , 120, 120, 100 100 1 so 80

60 80

40 40

20 20

0 0 3ClV WC,” 15OCIV COillTOlS 3av 30 a v 15oav COOtlds 3 CI v 30 CI v mav

Figure 4. Progressive neuronal loss induced in limbic structures by repeated brief kindled seizures. The bars represent the average neuronal densities of the septal and temporal poles ipsilateral and contralateral to the stimulating electrode at each location, expressed as a percentage of the neuronal densities at equivalent locations in unstimulated electrode-implanted control rats. Significant differences (0.001 < p < 0.05) in the average neuronal densities are indicated by the stars. There were also significant reductions in the neuronal density in temporal regions of CAla and CA3a after three class V seizures, which are not reflected in the neuronal density calculated by averaging the septal and temporal densities (see text and Table 7 for details).

DG after 3 and 30 class V seizures (Cavazos and Sutula, 1990). able, followed by neurons in CAl, CA3, EN, EC, granule cells In the present study, the hilus of the DC and the temporal poles in the DG, and CA2. of CA lc, CAla, and CA3a were particularly vulnerable to sei- Significant differences in neuronal densities as a function of zure-induced neuronal loss. Neuronal loss was observed in these stimulation site were not observed after 3, 30, or 150 class V locations after as few as three class V seizures. At this time seizures. However, the methods of this study do not exclude point, no other limbic structures evaluated in this study exhib- the possibility that at earlier stages of kindling there might be ited significant neuronal loss. subtle differences in the regional patterns of neuronal loss that After 30 class V seizures, significant decreases in neuronal could reflect patterns of seizure propagation dependent on the density were observed in the pyramidal layer of CA3c, CA3a, specific circuitry activated by local stimulation of the perforant CAlc, CAla, layers 2 and 6 of the lateral EC, layer 2 of the path, amygdala, or olfactory bulb. medial EC, and the rostra1 EN. After 150 class V seizures, sig- In the hilus of the DG, the loss of polymorphic neurons was nificant decreases in neuronal density were also observed in layer initially quite abrupt, with a 20% decrease in neuronal density 6 of the medial EC, in the temporal stratum granulosum of the after only three class V seizures. The decrease in neuronal den- DG, and possibly also in temporal CA2. Although the statistical sity in the hilus reached 35% after 30 class V seizures, and was analysis included corrections for multiple comparisons, the sig- 48% after 150 class V seizures. In comparison, the onset of nificance level of the decrease in neuronal density in CA2 (p < detectable neuronal loss was slower and of lesser magnitude in 0.03) must be interpreted with caution. Nevertheless, the overall other limbic structures. This observation is consistent with pre- results demonstrated a hierarchy of selective vulnerability to vious evidence for a sensitive subpopulation of neurons in the seizure-induced damage in the limbic system, which progressed hilus, which lack the calcium-binding proteins parvalbumin and in a variety of locations in parallel with the continuing evolution calbindin and are exquisitely vulnerable to continuous seizures of kindling. Neurons in the hilus of the DG were most vulner- (status epilepticus) (Sloviter, 1989). 3116 Cavazos et al. * Neuronal Loss Induced by Kindling

Figure 5. Neuronal loss induced in the DC by repeated brief kindled seizures. A, Higher magnification of the horizontal cresyl violet-stained section of the dorsal DG from the normal rat in Figure 1 demonstrates the normal distribution of neurons in the hilus and stratum granulosum at the septal pole of the hippocampal formation. B, Horizontal section at an equivalent location from a rat that experienced 150 generalized tonic- clonic (class V) seizures induced by repeated kindling stimulation of the perforant path. There appears to be a reduction in the population of polymorphic neurons; stereological methods revealed an average 49% decrease in the polymorphic neurons at this site compared to controls. The neuronal density in the stratum granulosum at this level was not significantly different than controls. C, Higher magnification of the horizontal cresyl violet-stained section of the temporal pole of the hippocampal formation from the normal rat in Figure 3 demonstrates the distribution of neurons in the stratum granulosum and hilus of the DG at this level. D, Horizontal section of the temporal DG at an equivalent location from the rat that experienced 150 generalized tonic-clonic (class V) seizures (same rat as in B). There appears to be a reduction in the polymorphic neurons in the hilus; stereological methods revealed an average 46% decrease of neurons in the hilus compared to controls. The stratum granulosum appeared normal to visual inspection, but stereological methods revealed an average 20% reduction in the neuronal density in the stratum granulosum compared to controls. DG, stratum granulosum of the DG; H, hilus. Scale bar, 200 pm. The Journal of Neuroscience, May 1994, 74(5) 3117

Figure 6. Neuronal loss induced in the stratum pyramidale of CAlc by repeated brief kindled seizures. A, Higher magnification of the horizontal section in Figure 2 demonstrates the normal distribution of neurons in the stratum pyramidale of CAlc at the septal pole of the hippocampal formation. B, Horizontal section at an equivalent location from a rat that had experienced 150 class V seizures (same rat as in Fig. 5&D). Stereological methods revealed an average 19% decrease in the neuronal density in septal CA lc compared to controls. C, Higher magnification of horizontal section in Figure 3 demonstrates the normal neuronal distribution of neurons in the stratum pyramidale of CA lc at the temporal pole of the hippocampal formation. D, A horizontal section of CA 1 c at an equivalent location from the rat that experienced 150 class V seizures (same rat as in B and Fig. 5B, D). Stereological methods revealed that rats kindled to 150 class V seizures had an average 38OYodecrease in the neuronal density in temporal CAlc compared to controls. Scale bar, 200 pm.

There is heterogeneity ofthe transmitter systems in the limbic immunocytochemical techniques could further identify the neu- circuitry that undergoes neuronal loss in response to repeated rons that are damaged by repeated kindled seizures (Kamphuis seizures. It is not possible by the methods of this study to de- et al., 1986). termine more specifically the identity of the seizure-sensitive The observation of neuronal loss in the rostra1 EN is also of neurons in the limbic populations that undergo neuronal loss interest, as recent observations have suggested that neural cir- during the evolution of kindling. Additional experiments using cuitry in this region may play an important role in the initiation 3118 Cavazos et al. * Neuronal Loss Induced by Kindling

Figure 7. Neuronal loss induced in the stratum pyramidale of CAla by repeated brief kindled seizures. A, Higher magnification of the horizontal section of the dorsal hippocampal formation in Figure 2 demonstrates the normal distribution of neurons in the pyramidal layer of CAla at the septal pole of the hippocampus. B, Horizontal section of the pyramidal layer of CAla at an equivalent location from a rat that experienced 150 class V seizures induced by repeated kindling stimulation of the perforant path (same rat as Figs. 5B,D; 6B,D). Stereological methods revealed that rats kindled to 150 class V seizures had an average 25% decrease in the neuronal density of the stratum pyramidale in septal CAla compared to controls. C, Higher magnification of the horizontal section in Figure 3 demonstrates the normal distribution of neurons in the stratum pyramidale of CAla at the temporal pole of the hippocampal formation. D, A horizontal section of the CAla region at an equivalent location from a rat that experienced 150 class V seizures (same rat as in B and Figs. 5B,D; 6B,D). Stereological methods revealed that rats kindled to 150 class V seizures had an average 44% decrease in the neuronal density of the stratum pyramidale of temporal CAla compared to controls. Scale bar, 200 pm.

ofepileptic activity in limbic pathways (Piredda and Gale, 1985, 1986; Racine et al., 1988; Hoffman and Haberly, 1991). Ad- Septotemporal variation of neuronal densities in normal rats ditional studies will be necessary to evaluate more comprehen- In this study, there were distinct differences in neuronal density sively the specific regions of this structure that may be affected along the septotemporal axis in the hippocampal formation of by seizure-induced neuronal loss. normal Sprague-Dawley rats. The neuronal densities in the hilus The Journal of Neuroscience, May 1994, 14(5) 3119

P

Figure 8. Neuronal density in layers 2 and 6 of the SSC in normal and kindled rats. A, Higher magnification of the horizontal section of the dorsal hippocampal formation and adjacent neocortex of a normal rat in Figure 1 demonstrates the distribution of neurons in the layer 2 of SSC. B, A horizontal section of layer 2 of the SSC at an equivalent location from a rat that experienced 150 class V seizures induced by repeated kindling stimulation of the perforant path (same rat as in Figs. 5&D; 6&D; 7&D). Stereological methods revealed that rats kindled to 150 class V seizures had a neuronal density that did not differ from controls. C, Higher magnification of the horizontal section of the normal rat in Figure 1 demonstrates the distribution of neurons in the layer 6 of SSC. D, A horizontal section of the layer 6 of the SSC at an equivalent location from the rat that had experienced 150 class V seizures (same rat as in B and Figs. 5&D; 6&D, 7&D). Stereological methods revealed that the neuronal density at this location in rats kindled to 150 class V seizures did not differ from controls. Z, IZ, V, and VI indicate the cortical layers. Scale bar, 200 pm.

of the DC, CA3a, and layer 6 of the medial EC were significantly onstrated significant differences in the anatomical organization higher in the temporal pole of the hippocampal formation than of the hippocampus along its septotemporal axis in normal an- in the septal pole. In contrast, the neuronal density in CAla imals (Amaral et al., 1990). For example, the ratio of baskets and layer 2 of the medial and lateral EC was significantly higher cells to granule cells in the DG is about 1: 140 in the septal pole in the septal pole than in the temporal pole. of the DG, but is only 1:220 in the temporal pole of the DG These observations confirm previous studies that have dem- (Seress and Pokorny, 198 1). In addition, the mossy fiber path- 3120 Cavazos et al. * Neuronal Loss Induced by Kindling

way does not project to the inner molecular layer in septal disturbances that are frequently encountered in temporal lobe (dorsal) DG in the normal rat, but does innervate the inner epilepsy (Sutula, 1990). molecular layer in the temporal (ventral) pole (Cavazos et al., For over a century, there has been controversy about hip- 1992). The functional effects of these differences in organization pocampal sclerosis as a cause or effect of the repeated, brief along the septotemporal axis are uncertain, but could influence seizures that are a feature of temporal lobe epilepsy (Babb and information processing and epileptogenesis in hippocampal cir- Brown, 1987; Gloor, 199 1). The observations of this study pro- cuitry. vide strong evidence that repeated brief seizures induce hip- There are variations in physiological properties along the sep- pocampal damage in a pattern that resembles hippocampal scle- totemporal axis. For example, there are differences in synaptic rosis, but do not exclude the possibility that neuronal loss and transmission along the septotemporal axis (Brazier, 1970; Ra- the associated sprouting, reactive synaptogenesis, and synaptic tine et al., 1977; Gilbert et al., 1985; Bragdon et al., 1986). The reorganization might also contribute to enhanced susceptibility relationship of these physiological differences to variations in to further seizures. The functional effects of neuronal loss, anatomical organization along the septotemporal axis is unclear, sprouting, alterations in inhibition, and the numerous cellular but could influence the development of epilepsy, as the temporal and molecular alterations in limbic circuitry associated with the pole is more vulnerable than the septal pole to the development development of seizures are extremely complex, and may vary of kindling (Racine et al., 1977). The observation of greater dynamically during the generation of seizures in limbic circuitry vulnerability to seizure-induced neuronal loss in the temporal reorganized by structural lesions and seizure-induced neuronal poles of the hilus of the DG and CAlc in rodents is also of loss (Sutula et al., 1992). This study has identified a variety of interest, as the anterior pole of the human hippocampus (which regions in limbic circuitry in which these complex issues might corresponds to the rodent temporal pole) is frequently involved be addressed in an effort to understand the cellular and molec- in human temporal lobe epilepsy, and may be more prominently ular processes that contribute to the disorder of temporal lobe involved in the lesion of hippocampal sclerosis (Babb et al., epilepsy. 1984). The present results are consistent with the possibility that excitotoxic mechanisms could contribute to neuronal damage Relationship of kindling-induced neuronal loss to hippocampal in the limbic system in association with repeated brief seizures. sclerosis The results provide a rationale for development of drugs that The lesion of hippocampal sclerosis, which is frequently ob- could reduce excitotoxic damage, and modify structural alter- served in human epilepsy, is characterized by neuronal loss and ations induced by seizures in limbic circuitry. Previous studies gliosis in CAl, CA3, and the hilus of the DG (Margerison and have demonstrated that brief treatment that suppresses intense Corsellis, 1966; Mouritzen Dam, 1980; Babb et al., 1984; Babb acute seizures in the kainic acid model of status epilepticus can and Brown, 1987; Gloor, 199 1). Recent morphological studies modify excitotoxic damage in the DG, prevent reactive mossy have demonstrated that this pattern of neuronal loss in humans fiber sprouting, and reduce susceptibility to chronic seizures with epilepsy is frequently associated with sprouting in the mossy (Sutula et al., 1992). Therapeutic efforts to reduce excitotoxic fiber pathway and other neurotransmitter systems, as well as damage after repeated brief seizures may also modify long-term dispersion of granule cells in the DG that may indicate devel- structural and functional alterations in limbic circuitry that con- opmental abnormalities (de Lanerolle et al., 1989; Represa et tribute to abnormal excitability in chronic epileptic models and al., 1989; Sutula et al., 1989; Houser et al., 1990; Houser, 199 1). human epileptic syndromes. In this study, repeated brief seizures evoked by kindling in- duced neuronal loss in the hippocampus and DG in a pattern References that resembled the principal features of hippocampal sclerosis. Amaral DG (1978) A Golgi study of cell types in the hilar region of In human epilepsy, neuronal loss may be observed in all sub- the hippocampus’in the rat. J Cdmp Neural 182:851-914. - regions of the hippocampus (Rim et al., 1990) and usually Amaral DG, Ishizuka N, Claibome B (1990) Neurons, numbers and affects the anterior pole most prominently, which corresponds the hippocampal network. Understanding the Brain Through the Hip- to the temporal pole in the rat. The observations that repeated pocampus 83: l-l 1. Babb T, Brown W (1987) Pathological findings in epilepsy. In: Surgical kindled seizures eventually induce neuronal loss in “resistant” treatment of the (Engel J, ed), pp 51 l-540. New York: neuronal populations such as granule cells in the DG and py- Raven. ramidal neurons in CA2, and more prominently reduces neu- Babb T, Pretorius W, Davenport C, Lieb J, Crandall P (1984) Tem- ronal populations in the temporal pole of the rat, thus have a poral lobe volumetric cell densities in temporal lobe epilepsy. Epi- remarkable parallel to human pathological observations. These lepsia 25721-732. Ben Ari Y (1985) Limbic seizure and brain damage produced by kainic observations suggest that kindling and human temporal lobe acid: mechanisms and relevance to human temporal lobe epilepsy. epilepsy could share common mechanisms, and also raise the Neuroscience 14:375403. possibility that kindling could play a role in the development Bertram E, Lothman E (1993) Morphometric effects of intermittent and natural history of temporal lobe epilepsy. kindled seizures and limbic status epilepticus in the dentate gyrus of the rat. Brain Res 603:25-3 1. Repeated kindled seizures in rats also induced neuronal loss Bragdon A, Taylor D, Wilson W (1986) Potassium-induced epilep- in other limbic structures such as the EC and the rostra1 EN. tiform activity in area CA3 varies markedly along the septotemporal These findings suggest that seizure-induced damage may not be axis of the rat hippocampus. Brain Res 378:169-173. limited to the hippocampus, and suggest the possibility that Brazier MA (1970) Regional activities within the human hippocampus and hippocampal gyms. Exp Neurol26:354-368. extrahippocampal limbic pathways may undergo structural re- Cavazos JE. Sutula T (1990) Progressive neuronal loss induced by organization as a result of repeated seizures. The neuronal loss kindling: a possible mechanism for mossy fiber synaptic reorgani- induced in hippocampal and limbic circuitry by brief repeated zation and hippocampal sclerosis. Brain Res 527: l-6. seizures could also play a role in the memory and cognitive Cavazos J, Sutula T (1991) loss in CA1 and CA3 The Journal of Neuroscience, May 1994. 14(5) 3121

induced by kindling: evidence that repeated sporadic seizures may Olney J, Rhee V, Ho 0 (1974) Kainic acid: a powerful neurotoxic cause hippocampal sclerosis. Epilepsia 32:66. analogue of glutamate. Brain Res 77:507-5 12. Cavazos J, Golarai G, Sutula T (199 1) Mossy fiber synaptic reorga- Piredda S, Gale K (1985) A crucial epileptogenic site in the deep nization induced by kindling: time course of development, progres- prepiriform cortex. Nature 3 17:623-625. sion, and permanence. J Neurosci 11:2795-2803. Piredda S, Gale K (1986) Role of excitatory amino acid transmission Cavazos JE, Golarai G, Sutula T (1992) Septotemporal variation of in the genesis of seizures elicited from the deep prepiriform cortex. the supragranular projection ofthe mossy fiber pathway in the dentate Brain Res 377:205-210. gyrus of normal and kindled rats. Hippocampus 2:363-372. Racine R (1972) Modification of seizure activity by electrical stimu- Choi D, Rothman S (1990) The role of glutamate neurotoxicity in lation. II. Motor seizure. Electroencephalogr Clin Neurophysiol 32: hypoxic-ischemic neuronal death. Annu Rev Neurosci 13: 17 l-l 82. 28 l-294. Clarke PC (1992) How inaccurate is the Abercrombie correction factor Racine R, Rose P, Bumham WM (1977) Afterdischarge thresholds for cell counts? Trends Neurosci 15:2 1 l-2 12. and kindling rates in dorsal and ventral hippocampus and dentate Crandall JE, Bernstein J, Boast C, Zometzer S (1979) Kindling in the gyrus. Can J Neurol Sci 4:273-278. rat hippocampus: absence of dendritic alterations. Behav Neural Biol Racine R, Mosher M, Kairiss E (1988) The role of the pyriform cortex 27:5 16-522. in the generation of interictal spikes in the kindled preparation. Brain de Lanerolle N, Kim J, Robbins R, Spencer D (1989) Hippocampal Res 454~251-263. interneuron loss and plasticity in human temporal lobe epilepsy. Brain Represa A, le Gall la Salle G, Ben-Ati Y (1989a) Hippocampal plas- Res 495:387-395. ticity in the kindling model of epilepsy in rats. Neurosci Lett 99:345- Gilbert M, Racine R, Smith GK (1985) Epileptiform burst responses 350. in ventral versus dorsal hippocampal slices. Brain Res 36 1:389-39 1. Represa A, Robain 0, Tremblay E, Ben-Ari Y (1989b) Hippocampal Gloor P (199 1) Mesial temporal sclerosis: historical background and plasticity in childhood epilepsy. Neurosci Lett 99:35 l-355. an overview from a modem prospective. In: Epilepsy surgery (Luders Seress L, Pokomy J (198 1) Structure of the granular layer of the rat H, ed), pp 689-703. New York: Raven. dentate gyrus: a light microscopic and Golgi study. J Anat 133: 18 l- Goddard G, McIntyre D, Leech C (1969) A permanent change in brain 195. function resulting from daily electrical stimulation. Exp Neurol 25: Sloviter R (1983) Epileptic brain damage in rats induced by sustained 295-330. electrical stimulation of the perforant path. I. Acute electrophysio- Hauser WA (1983) Status epilepticus: frequency, etiology, and neu- logical and light microscopic studies. Brain Res Bull 10:675-697. rological sequelae. Adv Neurol 34:3-14. Sloviter R (1987) Decreased hippocampal inhibition and a selective Hoffman W, Haberly L (199 1) Bursting-induced epileptiform EPSPs loss of interneurons in experimental epilepsy. Science 235:73-75. in slices of piriform cortex are generated by deep cells. J Neurosci 11: Sloviter R (1989) Calcium-binding protein (calbindin-D28) and par- 2021-2031. valbumin immunocytochemistry: localization in the rat hippocampus Houser CR (1991) Granule cell dispersion in the dentate gyrus of with specific reference to the selective vulnerability of hippocampal humans with temporal lobe epilepsy. Brain Res 535:195-204. neurons to seizure activity. J Comp Neurol 280: 183-l 96. Houser C, Miyashiro J, Swartz B, Walsh G, Rich J, Delgado-Escueta Sloviter R, Damiano B (198 1) Sustained electrical stimulation of the AV (1990) Altered patterns of dynorphin immunoreactivity suggest perforant path duplicates kainate-induced electrophysiological effects mossy fiber reorganization in human hippocampal epilepsy. J Neu- and hippocampal damage in rats. Neurosci Lett 24:279-284. rosci 10:276-282. Stringer JL, Lothman EW (1992) Bilateral maximal dentate activation Kamphuis W, Wadman W, Buijs RM, Lopes da Silva FH (1986) De- is critical for the appearance of an afterdischarge in the dentate gyrus. crease in number of hippocampal gamma-aminobutyric acid (GABA) Neuroscience 46:309-314. immunoreactive cells in the rat kindling model of epilepsy. Exp Brain Sutula T (1990) Experimental models of temporal lobe epilepsy-new Res 64:49 l-495. insights from the study of kindling and synaptic reorganization. Epi- Kim J, Guimaraes PO, Shen MY, Masukawa LM, Spencer D (1990) lepsia 3 1:45-54. Hippocampal neuronal density in temporal lobe epilepsy with and Sutula T, Steward 0 (1986) Quantitative analysis of synaptic poten- without gliomas. Acta Neuropathol (Berl) 80:41-45. tiation during kindling of the perforant path. J Neurophysiol56:732- Konigsmark B (1969) Methods for the counting of neurons. In: Con- 745. temporary research methods in neuroanatomy (Nauta W, Ebbesson Sutula T, He XX, Cavazos J, Scott G (1988) Synaptic reorganization S, eds), pp 3 15-340. Berlin: Springer. in the hippocampus induced by abnormal functional activity. Science Margerison JH, Corselhs J (1966) Epilepsy and the temporal lobes. 239:1147-l 150. Brain 89:499-530. Sutula T, Cascino G, Cavazos J, Parada I, Ramirez L (1989) Mossy McNamara J (1986) The kindling model of epilepsy. Adv Neurol 44: fiber synaptic reorganization in the epileptic human temporal lobe. 303-318. Ann Neurol 26:321-330. Meldrum B, Vigouroux R, Brierley J (1973) Systemic factors and Sutula T, Golarai G, Cavazos J (1992a) Assessing the functional sig- epileptic brain damage: prolonged seizures in paralyzed artificially nificance of mossy fiber sprouting. Epilepsy Res [Suppl] 7:25 l-259. ventilated baboons. Arch Neural Psycho1 29:82-87. Sutula T, Cavazos J, Golarai G (1992b) Alteration of long-lasting Mouritzen Dam A (1980) Epilepsy and neuron loss in the hippocam- structural and functional effects of kainic acid in the hippocampus by pus. Epilepsia 21:617-629.- brief treatment with phenobarbital. J Neurosci 12:4 1734 187. Nadler J. Cuthbertson G (1980) Kainic acid neurotoxicitv toward the West M, Danscher G, Gydesen H (1978) A determination of the hippodampal formation:‘dependence on specific excitatory pathways. volumes of the layers of the rat hippocampal region. Cell Tissue Res Brain Res 195:47-56. 188:345-359.